In the past, new leads for drug discovery have been generated by random cross screening of single synthetic compounds made individually in the laboratory or by screening extracts obtained from natural product sources such as microbial metabolites, marine sponges and plants. A second approach has been rational drug design based on the structure of known biologically active compounds and/or their sites of biological action. This has now been complemented by the powerful techniques of computer-assisted drug design.
There has recently been an explosion in the availablity of new screening targets arising from the output of efforts to sequence the human genome and bacterial genomes. This has led to the development of high throughput screening techniques. Groups of compounds, typically eight, are exposed to a biological target. These groups may be assembled from collections of compounds previously individually prepared and since stored in a compound bank, the assembly being random or guided by the use of "similarity" programs. In addition, there has also been a rapid growth in the deliberate preparation and use of so-called libraries and/or arrays of compounds. Each library contains a large number of compounds which are screened against a biological target such as an enzyme or a receptor. When a biological "hit" is found, the compound responsible for the "hit" is identified. Such compound, or lead, generally exhibits relatively weak activity in the screen but forms the basis for the conduct of a more traditional medicinal chemistry program to enhance activity. The libraries may be prepared using the rapidly developing techniques of combinatorial chemistry or by parallel synthesis (DeWitt et al, Proc Natl Acad Sci, 90, 6909, August 1993; Jung et al, Angew Chem Int Ed Engl, 31:367-83, 1992; Pavia et al., Bioorg Med Chem Lett, 3:387-1993).
The first libraries were composed of small polypeptides, with some libraries containing up to 10,000 members. Such libraries could be made by adapting the techniques developed for the synthesis of single polypeptides (see, for instance, Lam et al, Nature, 354: 82, 1991 and WO 92/00091; Geysen et al, J Immunol Meth, 102: 259, 1987; Houghten et al, Nature, 354: 84, 1991 and WO 92/09300 and Lebi et al, Int J Pept Prot Res, 41, 201, 1993). The chemistry involved, forming an amide bond, is relatively straightforward and automated peptide synthesisers can be employed to reduce the manual effort involved. However, small polypeptides do not provide ideal leads for drug discovery. Peptides are not generally useful as therapeutic agents and there exists no rational way of translating a peptide into a therapeutically useful small molecule (peptidomimetic). A similar approach has also been used with nucleotides, taking advantage of the progress made in automated nucleotide synthesis, and with oligomers.
Attention has therefore turned to preparing libraries of small non-peptide molecules based upon a common template or core structure see for instance Ellman and Bunin, J Amer Chem Soc, 114:10997, 1992 (benzodiazepine template), WO 95/32184 (oxazolone and aminidine template), WO 95/30642 (dihydrobenzopyran template) and WO 95/35278 (pyrrolidine template)!. The template will have a number of functional sites, for instance three, each of which can be reacted, in a step-wise fashion, with a number of different reagents, for instance five, to introduce 5.times.5.times.5 different combinations of substituents, giving a library containing 125 components. The library will normally contain all or substantially all possible permutations of the substituents. The template may be a so-called `biased` template, for instance incorporating a known pharmacophore such as a benzodiazepine ring or a so-called `unbiased` template, the choice of which is influenced more by chemical than biological considerations. Unbiased templates are considered to offer the greater potential for generating entirely new leads.
The real challenge in creating a small molecule library which is useful as a screening tool is to provide a diverse range of substituents comprising a wide range and variety of structural units which allow the library as a whole to explore as fully as possible the active site of a receptor or an enzyme in an assay by having the potential for a wide range of interactions such as hydrogen bonds, salt bridges, .pi.-complexation, hydrophobic effects etc. The actual substituents are selected by considering their physico-chemical properties such as, for example, electronic, ionic, lipophilic and steric properties in order that the library contains maximum structural diversity. For example, if a core structure is to have a C.sub.1-6 alkyl substituent at a particular position, a typical library may have component compounds in which that substituent is methyl and t-butyl. An adamantyl group provides a good example of a large, bulky hydrophobic group. Substituents on an aromatic ring may be varied according to well established principles of medicinal chemistry, e.g., as reflected in the Topliss and Craig diagrams. Suitable diverse heteroaryl groups may be chosen according to well-known medicinal chemistry principles. For instance, a pyridinyl group may be selected if a basic group is desired. In addition, computer programs have now been developed to assist in this process, for instance SYBYL molecular diversity manager (Tripos Inc, Mo, USA). It is also useful to avoid mass redundancies when selecting suitable substituents, to aid identification of different library members by mass spectroscopy. Tables have been devised to assist in this task (PCT/EP96/03731, SmithKline Beecham).
For maximum synthetic efficiency in creating a library, the introduction of different substituents at each functional site should be accomplished as a single step, using a mixture of reagents, one for each different substituent. A diverse range of substituents can however translate into a diverse range of reactivities for the reagents. It is often more convenient to adopt the so-called `split and mix` approach whereby the evolving library is split into a series of parallel aliquots, each containing the same mixture. Each aliquot is then reacted with a single but different reagent, to introduce a further variant, and the new sub-libraries can then be recombined before splitting again, for a further synthetic cycle (Furka et al, 14th Intl Congress of Biochemistry, Prague, July 1988; Furka et al, Int J Peptide Protein Res, 37: 487, 1991). Such an approach is of assistance in coping with different reactivities of diverse reagents and also in deconvoluting a library, once a hit is found. The progress of reactions may be monitored using various techniques, for instance the disappearance of a functional group such as an amine. Single bead mass spectroscopy allows the possibility of selectively sampling and analyzing large numbers of compounds, enabling this technique to be used to monitor and/or analyze libraries. Solid phase NMR, in particular so-called `magic angle` NMR, can also be usefully applied.
A complementary approach to creating a library of compounds is to use the parallel synthesis method, whereby the compounds comprising the library are prepared separately and in parallel. Usually, the various reaction steps are not monitored and little or no effort is made to purify or isolate intermediate compounds (DeWitt et al, Proc Nat Acad Sci USA, 90:6909-13, 1993). The chemistry may be carried out in the solution phase or using solid phase supports. This allows for a greater rate of synthesis, although at the possible expense of incomplete reactions. Compounds may be screened individually or they may be grouped together, for instance if there is a limited supply of screening target. Either way, deconvolution, once a "hit" is found, is then assisted by the existence of individual compounds. This approach is becoming increasingly automated and is attractive for the preparation of a small number of compounds. For larger libraries, the combinatorial approach becomes increasingly more efficient, as far fewer reactions have to be carried out.
The screens in which the libraries are assayed tend to be based on enzymes or receptors. These are becoming increasingly automated, giving them a high throughput and making the use of libraries more attractive. Furthermore, once created, libraries can become a screening resource which can be used many times over, both for existing screens and, held in reserve, for new screens as they are developed.